Background Africa bears 24% of the global burden of disease but has only 3% of the world's health workers. Substantial variation in health worker performance adds to the negative impact of this significant shortfall. We therefore sought to identify interventions implemented in sub-Saharan African aiming to improve health worker performance and the contextual factors likely to influence local effectiveness. Methods and Findings A systematic search for randomised controlled trials of interventions to improve health worker performance undertaken in sub-Saharan Africa identified 41 eligible trials. Data were extracted to define the interventions' components, calculate the absolute improvement in performance achieved, and document the likelihood of bias. Within-study variability in effect was extracted where reported. Statements about contextual factors likely to have modified effect were subjected to thematic analysis. Interventions to improve health worker performance can be very effective. Two of the three trials assessing mortality impact showed significant reductions in death rates (age5 case fatality 5% versus 10%, p0.01; maternal in-hospital mortality 6.8/1000 versus 10.3/1000; p0.05). Eight of twelve trials focusing on prescribing had a statistically significant positive effect, achieving an absolute improvement varying from 9% to 48%. However, reported range of improvement between centres within trials varied substantially, in many cases exceeding the mean effect. Nine contextual themes were identified as modifiers of intervention effect across studies; most frequently cited were supply-line failures, inadequate supervision or management, and failure to follow-up training interventions with ongoing support, in addition to staff turnover. Conclusions Interventions to improve performance of existing staff and service quality have the potential to improve patient care in underserved settings. But in order to implement interventions effectively, policy makers need to understand and address the contextual factors which can contribute to differences in local effect. Researchers therefore must recognise the importance of reporting how context may modify effect size.